Scholarlink

<p>This study presents an integrated framework that combines Recurrent Neural Networks (RNNs) with geostatistical modeling to enhance permeability prediction and spatial characterization of subsurface reservoirs. Using Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) architectures, the approach captures complex nonlinear, depth-dependent relationships within well log data. At the same time, subsequent geostatistical modeling translates these predictions into three-dimensional reservoir models. Applied to the Upper Surmeh Formation in the Persian Gulf, the integrated workflow demonstrates improved accuracy and spatial consistency compared to conventional deterministic methods. Stratigraphically equivalent to the Arab Formation, the Upper Surmeh Formation comprises evaporitic and carbonate facies in the study area. This study introduces an integrated framework that combines Long Short-Term Memory (LSTM) neural networks with geostatistical modeling to enhance reservoir permeability prediction and spatial characterization. Compared to traditional machine learning models such as Random Forest (RF), Support Vector Regression (SVR), and Gradient Boosting Regression (GBR), the LSTM approach achieved superior performance (MAE = 0.078, RMSE = 0.13). The key innovation of this research lies in embedding LSTM-derived permeability predictions into three-dimensional Sequential Indicator Simulation (SIS), enabling geologically consistent and spatially realistic reservoir models. Applied to the Upper Surmeh Formation in the Persian Gulf, the integrated deep learning–geostatistical workflow effectively captures nonlinear depth-dependent patterns and lateral heterogeneity. The results demonstrate that this hybrid approach significantly improves prediction accuracy and geological realism, providing a robust and scalable methodology for reservoir characterization in heterogeneous carbonate–evaporitic systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integration of the RNN and geostatistics methods for permeability prediction: a case study from the surmeh formation, Persian Gulf

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月19日

查看全文
  • Mehran Rahimi,
  • Mohammad Ali Riahi

摘要

This study presents an integrated framework that combines Recurrent Neural Networks (RNNs) with geostatistical modeling to enhance permeability prediction and spatial characterization of subsurface reservoirs. Using Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) architectures, the approach captures complex nonlinear, depth-dependent relationships within well log data. At the same time, subsequent geostatistical modeling translates these predictions into three-dimensional reservoir models. Applied to the Upper Surmeh Formation in the Persian Gulf, the integrated workflow demonstrates improved accuracy and spatial consistency compared to conventional deterministic methods. Stratigraphically equivalent to the Arab Formation, the Upper Surmeh Formation comprises evaporitic and carbonate facies in the study area. This study introduces an integrated framework that combines Long Short-Term Memory (LSTM) neural networks with geostatistical modeling to enhance reservoir permeability prediction and spatial characterization. Compared to traditional machine learning models such as Random Forest (RF), Support Vector Regression (SVR), and Gradient Boosting Regression (GBR), the LSTM approach achieved superior performance (MAE = 0.078, RMSE = 0.13). The key innovation of this research lies in embedding LSTM-derived permeability predictions into three-dimensional Sequential Indicator Simulation (SIS), enabling geologically consistent and spatially realistic reservoir models. Applied to the Upper Surmeh Formation in the Persian Gulf, the integrated deep learning–geostatistical workflow effectively captures nonlinear depth-dependent patterns and lateral heterogeneity. The results demonstrate that this hybrid approach significantly improves prediction accuracy and geological realism, providing a robust and scalable methodology for reservoir characterization in heterogeneous carbonate–evaporitic systems.